[solved] VM network driver MTU size problems

[Maor]

I have UNRAID 6.6.5 and setup a VM with Xubuntu 18.04 (kernel 4.15.0-20)

When I try to use apt It get stuck with the following:

sudo apt-get update
0% [Waiting for headers] [Waiting for headers]

I found out, that it is due to problem with MTU size. When the MTU of the "enp2s0" virtual netwotk interface is set to 1500, it simply does not work. When I set it to MTU = 1472 or similar smaller value it works.

 

Can someone please explain what is wrong? I never had to touch MTU size before, my Unraid server has MTU = 1500 and I never had a problem.

 

EDIT: I managed to hit similar problem with one of my Windows 10 VMs. After reboot the network stopped working, ping was ok. Adapter reseting did not work, only changing MTU to 1350 did the job. 

UPDATE: cause was Hamachi docker from gfjardim. Stopping using docker solved my problem.

Edited May 13, 2020 by Maor
update

[Maor]

Some time and some more linux experience and tcpdump dubugging and here is an update:

 

So I believe, that I found what is causing the issue:

My domains share has 16 VM of size 1,5TB.

 

When I change the default VM storage path to "domains2" which is emty, and delete the libvirt.img the packet fragmentation works.

I tried also to leave the default VM storage path to "domains" and delete thr libvirt.img and let it rebuild, but any newly created VM has the bug with packet fragmentation.

 

I believe, that by deleting vdiscs from "domains" I can isolate the files, which are causing the problem.

If anyone is interested, I can send my diagnostics.

[Maor]

So after some tinkering, aparently the problem is caused by gfjardim's Hamachi docker connected to br0. In my opinion that should not happen as the dockers are designed not to interfere with the rest of the system.
I will leave hamachi in favour to Wireguard.

Edited May 4, 2020 by Maor